Kibana CVE-2026-72675

Exploitation requires a valid Kibana user account with at least low-privilege access - confirmed by the CVSS PR:L metric. The deployment must use Kibana Spaces with multiple spaces configured, and the Machine Learning feature must be active and accessible to the authenticated user. The flaw is in specific ML API code paths that omit the per-request space filter; only those paths are affected. Exploitation is limited to authenticated users and does not apply to single-space Kibana deployments or to deployments where ML is not enabled or not accessible to lower-privilege roles. Remote unauthenticated exploitation is not possible.

CVSS 7.1 (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N) accurately reflects the risk profile: network-reachable, low complexity, but gated behind authentication at the low-privilege level. The High confidentiality impact is the dominant concern - cross-space ML data exposure can include model configurations, anomaly detection results, trained models, and associated index data that may contain sensitive business or security telemetry. The Low integrity score reflects limited but real write-side risk. No EPSS data was provided, and the vulnerability is not listed in CISA KEV, indicating no confirmed active exploitation at time of analysis. However, the low attack complexity and minimal privilege requirement mean any Kibana user account (including compromised service accounts or shared demo credentials) is sufficient to trigger the flaw. In multi-tenant environments or deployments where Kibana spaces are used to enforce data separation between teams or customers, this represents a meaningful insider threat and lateral information access risk. Risk is bounded by the requirement for valid Kibana authentication and only impacts deployments actually using ML features across multiple spaces.

An attacker who has obtained or been granted a low-privilege Kibana user account assigned to a single space (e.g., a shared analyst account or a compromised service token) issues Machine Learning API requests through the affected code paths. Because the space filter is not applied, the ML subsystem executes these requests against Elasticsearch using its elevated internal credentials without scope restriction, returning ML model data, anomaly records, and inference results from every space in the deployment. No public exploit code has been identified at time of analysis.